 Live from Las Vegas, Nevada, it's theCUBE. Covering IBM World of Watson 2016. Brought to you by IBM. Now, here are your hosts, John Furrier and Dave Vellante. Hey, welcome back everyone. We're live here in Las Vegas Mandalay Bay Convention Center with IBM World of Watson. This is Silicon Angles theCUBE, our flagship program where we go out to the events and extract the signal from the noise. I'm John Furrier, my co-host Dave Vellante, head of research at wikibond.com. Our next guest is Mary Glacken, who's the senior vice president of VP, senior vice president of science and forecast operations at the weather company. So, super excited. What a sense of a great job. I'll see that weather data we always geek out on. Certainly, we saw the catastrophe of the hurricane that could have been worse this past month. So, very, very timely conversation. Welcome to theCUBE. Thank you, thank you. Thrilled to be here. So obviously IBM is getting into the weather business. But you see it play out on internet of things. You're seeing it play out as an interesting acquisition because now weather factors into a lot of things. And something that's near indeed our heart is autonomous vehicles. Absolutely. Which is an area you don't really kind of correlate weather with cars. How does that relate? I mean, how does? Well, it has incredible relationships there. First, unfortunately, almost an extremely large percentage if I'm recalling correctly about 70% of accidents are weather related. So that's an attention gather right there. Seven zeros. Yes, seven zero. I think that's about right. I haven't looked at that recently. That's a big number. It's a big number. It's a big number. And we're excited about connected cars for connected vehicles really for a number of things. One is each of these vehicles gives us environmental information. So think about how your windshield wiper, how fast it's going, a whole set of that. We know something about how quickly the rain is falling. Think about every time your ABS thing engages there. We know something about the road surface there, whether it's wet or cold. We're obviously getting temperature information off of there. But what we want to do at the Weather Company IBM is what we're already doing in aviation is treating these vehicles as a two-way street. Extract that environmental information off of them and then provide them back a precise, accurate forecast. So in aviation today, we do this for turbulence forecasting. We take information off of the airplane and with our partner, Go-Go, we provide them back a turbulence alert. So as the planes are flying down, you can warn everybody behind you like that. So I think the internet of things. The trade-off on my Wi-Fi for checking email and streaming movies for a more comfortable ride. Absolutely. A safer ride, especially for those people that don't keep their seatbelt on. Exactly. Unless we're a terrible. I couldn't resist. I'm always complaining about how slow the Wi-Fi is. Then it's satellite-based. So it's going to land, it's going to ground. Okay, so that's really using precise measurement of the instrumentation of the plane. Exactly. And then you've got other data about the past. We are so excited about data. So if you think about weather, there's all those historical data. We've had observations at airports since we've had airplanes to do that. But you're seeing this fill in at the Weather Company. We have a network of private weather stations. These are weather enthusiasts like yourself that have purchased a weather station, have it in their backyard. Over 200,000, and most of our growth is global. So you're starting to see it go into areas where we've had a lack of observations like India, parts of Africa. Super exciting to see that. And in other traditional sources like satellites, we have the great big, I like to call them the sentinel satellites that the governments launch, but now we have all these CubeSats coming into play that are bringing really interesting environmental information. It's really a pretext to the trend that's happening because what's going on is that you're using crowdsource-like gamification of weather stations. Getting smaller and faster and cheaper with devices. And you're adding that into your existing system. That's right, and I love your tagline of extracting the signal from the noise. That's exactly what we're doing. If you have a lot of data like that, it's not going to be the highest quality data, but there's a signal in it, and that's what we extract. And you know what to look for too, because unlike they have one stream, you have other data that could tell you about that data. Exactly. So Mary, how does it work? You got this distributed device network that's collecting all this weather data. You got models. How much of the data actually goes back to the models? How much stays at the edge? Do you send the model to the data? How does it all work? Okay, so if you're going to run any model, you'd like to get the best set of highly quality control data to make it run. And we run a model at the weather company that we do, and we do it very custom for particular uses. So I'll give you an example over the northeast corridor of the US, that area between DC up to Boston that's so heavy, including New York City. We'll run a custom model for them with maximum information at a micro scale, really small scale. But what we really do to do our global solution is we take forecast models from all around the globe. We have about 160 of these. So we take the model output from the National Weather Service in this country, from the UK's Met Office, from the European Center, from the Bureau of Meteorology in Australia. We take all of those, and with a machine learning algorithm, we produce the best forecast. And we do it time and time again. You can see, we can be your best model any day of the week by machine learning on top of all of the models. So you begin to understand the biases of each of those? Absolutely, in real time. So if your model isn't tracking this storm particularly well, you're getting less weight in our, the algorithm understands that machine learning is adjusting on the fly. And the vast majority of the data comes back to the model, or is it sort of stay in place when you bring the model to the data? Well, for us, we bring model output with that, but model output is tremendously dense. You think of a model that's run every eight kilometers all around the globe. Think of time steps that are at least hourly, if not finer resolution. Think of it going out to 14 days in time. An incredible amount of data comes in and flows back and forth. I was mentioning earlier, we take in 100 terabytes of data a day at the weather company. We're ingesting that kind of data. We're talking physically, speed of light, coming into a central location, and then you apply the model there, okay? That's a lot of ingestion. That's a lot of ingestion. You get a lot of indigestion of the data. But that's interesting. So I want to ask you about that. So obviously, I want to, the ingestion's key, but one of the things that's coming up in the database world, this comes up in all of our shows we go to, you have database guys who are trained computer science guys who know schemas. Oh yeah, they can handle unstructured data. But what's going on now is that they ingest everything and think they're going to figure it out. But if you think about real time, weather's real time, time series data is not really good anymore after the time it's needed. So maybe think about staging that differently. So there's new practices emerging with time series specifically. What are you seeing there? Because you guys are dealing with this every day. You don't want to swallow the basketball, be that snake if you will, when you don't need it. So John, that's such a perfect question. You've really teed it up. So IBM bought us, I would say, for a couple of reasons. One is because weather impacts virtually every business. Number two is we have a great forecast. But number three, we're completely running on the cloud. And so we manage all of this data in the cloud coming and going. We have a data platform that we're now in the process of commercializing to do that because we've kind of solved that problem. Real time transactions. I was just talking to a customer earlier, running a nationwide set of golf courses has to have a really good forecast for the golf course. We're exposing our data stream to them and they said they've never had such an easy integration. We gave them the APIs they were up and running the next day. And what's there? Obviously, besides customer satisfaction, is there concern around like lightning obviously? Golfers don't want to get struck by lightning. What's the issue with golf courses? So a couple things. Safety is one issue, but it's about optimizing. And in that case, if you think about a golf course, you want to really minimize the number of chemicals you use but you can't afford to let pests come in. You can't have blighted areas in there. So they're running. We expose a couple of variables to them for temperature, precip and all. They know when certain pests will breed and they know when to do apply chemicals. So we're helping them. You're a cognitive greenskeeper, basically for them. I think you should think about weather forecasting is yesterday's metric used to be to make a great forecast. Today and tomorrow's metric is to make a great decision. So it's not good enough to have the right forecast. You have to be helping people make the right decision and that's what we do all the time. Well, Ginny Romney says in the commercial, a lot of the data's hidden. Bob Pichino has dark data. I'm not a big fan of dark data. It reminds me of the London Valley of the Web, but dark meaning, invisible. But that being said, you can only do forecast. That was your best outcome. Now you're saying, okay, that's table stakes. What's the decision that you make with the known data? So what we really want to do, going back to an airline example, so we want to tell them how much fuel to put on that plane because so hair is liable to be in a ground stop for a certain period of time and they're going to have a slower ingest rate. So we're jumping beyond what's the ceiling forecast, what's the visibility forecast, right to the impact and giving you the forecast. That's a great example of the plane one. I talked to the vice chairman of United Airlines on a panel I did with G.E. one time and he said literally they save over a billion dollars hard cash using IOT and data just on how to move patrolling them around for the fuels. Absolutely. You're talking about something specific where there's potentially life savings, but more importantly, it's real money. It's optimization. I mean, it's real cash that they save. It absolutely is and it touches areas you wouldn't think about. In the consumer part of our business, we can tell advertisers when to advertise and I'm not talking about snow shovels and snowstorms. I'm talking about when people buy more yogurt because they buy it under certain meteorological conditions. Very nuanced. Very nuanced and we have all of those insights. That's the signal from the noise. So you say that it's all about ultimately making the good decision, the right decision and it's not just about having accurate forecasts but that is, as John said, table stakes. And people like to joke about the weather. During that whole deflate gate thing, Bill Belichick was there. The weather forecast is terrible. They pile on. What is your data show in terms of the accuracy of a forecast over the last, whatever, in years, pick yours. So our overall forecast accuracy, if you pick a measure like temperature and getting that right within a degree, we're in the mid-80s, 80% to do that now. But what we really want to look- But the weather stage is now coming online. You're going to probably get more precise. Yes, we're getting more precise in that. I can imagine. But let me tell you one of the things we're really focused on because this is what's so important to so many of our users. It's not really what that temperature forecast is but when I'm going back to the gentleman I was talking about with the golf courses, he really wants to know when he's going to get a quarter of an inch or more of precipitation because that quarter of an inch makes a difference to him in terms of what he has to do for irrigation. So he doesn't really care whether it's going to rain or not. He cares how much it's going to rain and he cares exactly about the amount of rain. And that's important. And how are you doing there? Well, I can't cite a visibility thing for that. But my point is- That's a new challenge that you're taking. Yeah, it's a new challenge. And what we will do- No, that's the impact of the customer. That's an example of the impact of the customer benefit. Well, I want to even take a level like that. We can make him a forecast for that quarter of an inch but we'll give him a confidence probability with it. And then he can weight that with his other factors. And then learn over time, improve his or her model. Exactly. Okay, so what I just learned is it's not, like you say, it's not just about temperature and is it going to rain? What should I wear? You're being presented with new business challenges that are weather related that may not even show up on our little weather app. Absolutely. And I think every business needs to think about both how they can capitalize on weather and how they can minimize the costs of weather. Because I think there's both sides of the coin there. Mayor, interesting background that you have. First of all, this is a great conversation. We can geek out on weather. But I want to ask a personal question. How did you get into this? Because you have a computer science degree- Right. From Maryland, I noticed that on your profile. And you also worked for the US Secretary? Well, I actually spent 20 years in the National Weather Service. And I have about 30 credits in meteorology as well. I have a, so I'm old, right? We can see my gray hair. So in the late 70s, it was all about these great big computers and weather forecasting. So there was this nice marriage between mathematics, computer science, and meteorology. In fact, in Maryland, it was all in the same hallway. But how I really got into that was as a child and being really being impressed with storms, we got caught in, including one tropical storm with no warning. You got caught in. Right. Really? Our elementary school was closed down at 11 o'clock and we were released with the winds blowing and my mother would tell you the nightmare story of trying to hold on to everybody. Because we take it for granted now. And that impacted you. There was no satellites, no weather satellites, watching things. Well, the North East is on the East Coast is one historic storm that wiped out everyone. So you got into that. We'd love to have you on the more on theCUBE. We just had Grace Hopper last week talking about women in computing and going back to the old, to how math and the contribution in the industry across the board. But what's interesting is the global warming question. I want to ask you about global warming because that's something that didn't come up on the debates I was kind of upset about, that last debate. Global warming, how is this going to evolve in your mind? Obviously, how real is it? I mean, you see the numbers. It's pretty staggering. It seems to be very real. Yeah. So it is real. There's just no question about that. And we have tried to shift the language to climate change because not everybody gets warm, you know? Climate change, yeah. But things are changing. I think we're going to see these impacts and I think we're going to see a lot of them play out in extreme events. So we know already, if you go back and look at Hurricane Sandy, Superstorm Sandy that came into the Northeast, we can tell, you know, the science has been done to talk about how much more extensive the damage was because the sea level has changed dramatically, you know, or noticeably I should say, not dramatically over this past century. Well, dramatically given the storm, given the flood in Lower Manhattan. So I think that any business that's assuming the last 10 years of weather looks like the next 10 years is making a mistake. You're going to see more extreme events. It's, you know, it's common. It didn't used to be common to see three inches of rain, what we saw in Louisiana, you know, a couple of weeks ago, I guess it's been a month or six weeks ago now. Very common. So people that are, businesses and communities that are impacted by this are asking the question because the question really is, how often am I going to see this quote 50 year event? And the answer is you're seeing it a lot more often. A lot more sooner and more frequency of those. Is there anything in the data that you're learning now with cloud computing and some of the, you mentioned high. So let me say, make one point that I wanted to get in here. Okay. That's the question about how cognitive and cloud computing and all are going to help weather forecasting get smarter. And one of the things that we're excited about in IBM is the opportunity to have Watson look at the reams, go back and look at our past reams of data and help us understand some patterns. So for example, we know today that a warming pattern in the Pacific, an unusual warming pattern in the Pacific results in a drought in the middle of the country. What other patterns haven't we noticed and could Watson find them for us? So go dig through the archives. Go dig through those archives and find out what the scientists haven't seen. How big is the archive? Well, it's big. I mean, how big does it go back? How many years? There's also text paper. Has it been digitized? I mean. Oh yes, all of that's been digitized. The federal agencies in this case know I worked for a long time. The climatic data center there has all of those archives available. And you know, you have really good data going back to post World War II. And I think it would be, you know, we're anxious to put Watson loose on that. Well, they can throw some compute at that too. So you can theoretically sequence genomes in five minutes now. So potentially have Watson just grow. Yeah, and help us kind of figure out too. Maybe there's some things we can do. I'm confident there is in the forecast for the week three and week four and think about how important that would be for supply chains. And there's a predictive element here as well. Absolutely. Past is not prologue and you talked about, you know, climate change. So that's another factor that has to be put together. My final question, Mary, for you is what's the coolest thing that you've seen that you've seen in the past couple of years? One, that you've seen observed in technology and weather. And two, the coolest thing that you've done. So I think the coolest thing I've observed, I think that some of the observing that's done from space is awesome. I'll put in a plug then. NOAA is about to launch a new geostationary satellite in two weeks and it's going to be a game changer in terms of the ability to look at Earth and it, you know. So I think that that technology of space base observing which lets you do the whole Earth is just awesome. The coolest thing I think I have done is stood on the Greenland ice sheet. Both the coolest thing and the most sobering thing is to stand on the Greenland ice sheet and listen to it cav off. So you talk about climate change, you could stand there and watch it. And within your line of sight, you could see where that ice sheet was in 1970 and how quickly it's eroding. That's moving, great. Mary, thanks so much. Great conversation. We could have gone 45 minutes talking about weather and all the coolness of the tech behind it, the intersection of technology and just impact the life. Thank you so much. Thanks, it was a lot of fun. Very glad to have you as the Senior Vice President of Science and Forecast Operations that the weather company, part of an IBM business, will be back with more weather and signal here in the Cube after this short break. I'm John Furrier with Dave Vellante. We'll be right back. Thank you.